AI-Powered Clinical Tool

Predict Stroke Risk
Before It Happens

Leveraging explainable machine learning on clinical data to provide transparent, real-time stroke risk assessments. Not just a prediction a clinical decision support tool.

130,000+

Annual stroke deaths in Nigeria

95%

Model accuracy achieved

5,110

Patients analyzed in dataset

4

ML models ensembled

Real-Time Analysis

Get instant stroke risk predictions as you input patient data. No waiting, no delays results in milliseconds.

Explainable AI (SHAP)

Every prediction comes with transparent explanations. See exactly which factors contribute to risk and by how much.

What-If Simulator

Adjust risk factors with interactive sliders and watch how lifestyle changes impact stroke risk in real-time.

Multi-Model Ensemble

Four machine learning models Random Forest, XGBoost, Logistic Regression, and Neural Network working together for reliable predictions.

Stroke Risk Assessment

Enter patient health data below. The risk gauge updates in real-time as you type. Hit "Analyze" for a full explainable report.

Patient Health Profile

Fill in all fields for the most accurate prediction

High blood pressure

Pre-existing condition

Normal: 70-100 mg/dL

Normal: 18.5-24.9

Live Risk Preview

Updates as you type. Click "Run Full Analysis" for detailed SHAP explanations.

Fill the form

Dataset Insights

Data Analytics

Explore the stroke prediction dataset with interactive visualizations. Understanding the data is key to trusting the model.

5,110

Total Patients

records analyzed

249

Stroke Cases

4.9% of dataset

4,861

No Stroke

95.1% of dataset

10

Features

clinical attributes

Stroke Rate by Age Group

Stroke incidence increases dramatically with age

Global Feature Importance

Which features the model relies on most (SHAP global values)

Gender Distribution
Female (2,994)
Male (2,115)
Other (1)
Smoking Status & Stroke Cases
Never: 90 strokes
Formerly: 70 strokes
Smokes: 42 strokes
Unknown: 47 strokes
Interactive Simulator

What-If Risk Reduction

Adjust modifiable risk factors and watch how lifestyle changes impact stroke risk in real-time.

Adjust Risk Factors

Scenario: A 65-year-old male smoker with hypertension and elevated glucose

220 mg/dL
50Normal (<120)High (>200)350
32
12Normal (18.5-24.9)Obese (>30)60
65 years
13055+65+100
Adjusted Risk Level

Risk Reduction

0.0%

0.0% → 0.0%

No Change
Knowledge Hub

Understanding Stroke

Knowledge is the first line of defense. Learn the warning signs, risk factors, and prevention strategies.

The FAST Warning Signs

F

Face Drooping

One side of the face may droop or become numb. Ask the person to smile is it uneven or lopsided?

A

Arm Weakness

Sudden numbness or weakness in one arm. Ask the person to raise both arms does one drift downward?

S

Speech Difficulty

Slurred speech or inability to repeat a simple sentence correctly. Is the person's speech confused or garbled?

T

Time to Call Emergency

If any of these signs are present, call emergency services immediately. Every minute counts time is brain tissue.

Emergency: Call 112 (Nigeria)

If you or someone around you shows any FAST signs, do not wait. Call emergency services immediately.

Key Risk Factors

High Blood Pressure

The leading cause of stroke. Manage with medication, diet, and exercise.

Heart Disease

Atrial fibrillation and other heart conditions can cause blood clots that lead to stroke.

Diabetes

Diabetics have 1.5x higher stroke risk. Blood sugar control is critical.

Smoking

Damages blood vessels and raises blood pressure. Quitting reduces risk by 50% within a year.

Prevention Strategies

1

Maintain a healthy blood pressure through regular monitoring and medication

2

Exercise at least 150 minutes per week brisk walking counts

3

Follow a balanced diet rich in fruits, vegetables, and whole grains

4

Quit smoking and limit alcohol consumption

5

Manage diabetes and maintain healthy cholesterol levels

6

Get regular health check-ups, especially after age 45

Technical Details

Model Methodology

Our approach from raw data to a deployable clinical tool.

1

Data Collection

Kaggle Stroke Prediction Dataset 5,110 patient records with 10 clinical features

2

Preprocessing

BMI imputation (median), one-hot encoding, SMOTE for class balancing (5% → 50% positive)

3

Model Training

4 models trained: Random Forest, XGBoost, Logistic Regression, Neural Network

4

Evaluation

Stratified 5-fold cross-validation, optimized for F1-score and ROC-AUC on imbalanced data

5

Explainability

SHAP values computed for model transparency every prediction is explainable

6

Deployment

FastAPI backend serving the model, Next.js frontend for real-time predictions